The Forward Speed Diffraction Problem in

advertisement
The Forward Speed Diffraction Problem in
Following Seas
by
Gregory E. Osborne
B.A., Chemistry
'Vanderbilt University (1986)
M.S., Engineering
University of New Orleans (1990)
Submitted to the Department of Ocean Engineering
in partial fulfillment of the requirements for the degree of
Ocean Engineer
at the
MASSACHUSETTS INSTITUTE OF TECHNOLOGY
June 1994
© Massachusetts Institute of Technology 1994. All rights reserved.
Author
................
Department of Ocean Engineering
April 29, 1994
....................
J. N. Newman
Professor of Naval Architecture
Thesis Supervisor
Certified by. ..
Acceptedby.....
4.-
A. Douglas Carmichael
Professor of Power Engineering
Chairma m, Departmental Committee on Graduate Students
MASSACHUSETTS
INSTITUTE
JUN 2 0 1994
LionnlrrSI'b
ARCHIVES
The Forward Speed Diffraction Problem in Following Seas
by
Gregory E. Osborne
Submitted to the Department of Ocean Engineering
on April 29, 1994, in partial fulfillment of the
requirements for the degree of
Ocean Engineer
Abstract
When a ship moves through an incident wave field at steady forward speed, it is
exposed to forces due to both the oncoming waves and the scattered waves created
by the body. The problem of determining the forces attributable to these waves is
known as the diffraction problem. Herein, this problem is studied in general along
with its particular extension to the special case of a ship travelling at steady forward
speed encountering following seas.
The theoretical model for the solution of this problem is based in the time domain
and uses linear potential flow theory to determine the forces in terms of impulseresponse functions. The case of following seas requires special consideration because
the incident wave field is defined with respect to a coordinate system fixed with the
ship. Such an incident wave field causes there to be an ambiguity when we try to
resolve the wave input. Hence, it is necessary to modify the theory to accommodate
this ambiguity.
Numerical results are calculated using the three-dimensional panel code TIMIT.
This code is a suite of programs developed at the Computational Hydrodynamics
Facility which solves the canonical problems for the velocity potentials and uses them
to determine the hydrodynamic quantities in terms of impulse-response functions.
Results are presented for a Wigley hull at zero speed and at forward speed in both
head seas and following seas.
Thesis Supervisor: J. N. Newman
Title: Professor of Naval Architecture
'The time has come,' the Walrus said,
'To talk of many things:
Of shoes - and ships - and sealing wax Of cabbages - and kings And why the sea is boiling hot And whether pigs have wings.'
- Lewis Carroll
All four winds together
Can't bring the world to me
Shadows hide the play of light
So much I want to see
Chase the light around the world
I want to look at life - In the available light
- Neil Peart
Acknowledgments
I would like to thank my family and friends for all of their support during my time
at MIT. I would especially like to thank my mother and father for their unwavering
support and encouragement throughout my formative college years.
I also wish to express my sincere appreciation for the guidance of Dr. J. N. Newman
in the research and preparation of this thesis. Likewise, I wish to thank Dr. Tom
Korsmeyer for the helpful discussions.
The research presented in this thesis was sponsored by the Office of Naval Research, contract number N000i4-90-J-1160.
Contents
1 Introduction
15
2 Mathematical Formulation of the Problem
19
2.1
Exact Statement of Initial Value Problem ...........
2.2
Linearization of the Forward Speed Problem .........
2.3 Scattered Potential Formulation ................
2.4
Diffraction Potential Formulation
...............
2.5 The Impulse-Response Function ................
2.6 The Forward Speed Haskind Relations in the Time Domain.
2.7
Special Case: Following Seas ..................
3 Development of the Numerical Scheme
19
... . 22
... . 24
... . 28
. ...
30
... . 31
... . 38
43
3.1 Panel Methods.
43
3.2 Discretization of the Integral Equation .................
45
3.3 Determination of the Incident Potential .................
47
3.4
Calculation of the Wave Elevation ....................
58
3.5
Simulation .................................
62
4 Numerical Results
5
....
65
4.1
Zero Speed Results ............................
67
4.2
Forward Speed Results: Head Seas
70
4.3
Forward Speed Results: Following Seas .................
...................
77
93
Conclusions
9
10
List of Figures
4-1 Definition for the incident wave angle heading
..............
66
4-2 Impulsive Incident Head Wave for the Case of F, = 0.0 and , = 7r. .
4-3
68
Heave exciting force impulse-response function for a Wigley hull travelling at F, = 0.0 in head seas .......................
69
4-4 Pitch exciting moment impulse-response function for a Wigley hull
travelling at F, = 0.0 in head seas. ...................
70
4-5 Wigley hull at F,= 0.0, /3= r, the heave exciting force amplitude.
.
4-6 Wigley hull at Fn = 0.0, P = r, the pitch exciting moment amplitude.
71
71
4-7 Wigley hull at F, = 0.3, /3 = r, the heave exciting force impulseresponse function ..............................
72
4-8 Wigley hull at F, = 0.3, /3 = r, the pitch exciting moment impulseresponse function ..............................
4-9
72
Comparison of heave exciting force impulse-response functions for a
Wigley hull travelling at different forward speeds in head seas ....
74
4-10 Comparison of pitch exciting moment impulse-response functions for a
Wigley hull travelling at different forward speeds in head seas
.....
74
4-11 Wigley hull at F, = 0.3, 3 = r, the heave exciting force amplitude.
4-12 Wigley hull at F, = 0.3,
.
= r, the pitch exciting moment amplitude.
4-13 Wigley hull at F, = 0.3, /3= r, the heave exciting force phase angle.
75
76
76
4-14 Wigley hull at F, = 0.3, 3 = r, the pitch exciting moment phase angle. 77
4-15 Example time histories from a simulation of a Wigley hull at F, = 0.3,
,/ = r.
From the top: incident wave elevation in the ship-fixed frame,
heave response,
pitch
response.
. . . . . . . . . . . . . . .
11
. .
.
78
4-16 Wigley hull at F, = 0.3, P = a', the heave response-implitude operator.
4-17 Wigley hull at F, = 0.3,
/
= 7r,the pitch response-amplitude operator.
78
79
4-18 Convergence properties of the heave exciting force impulse-response
function for a Wigley hull travelling at F, = 0.3 in following seas
where <w
iw
-- .
2Ucosp'
. . . . . . . . . . . .. 80
. . . . . . . . . . . . . .80
4-19 Convergence properties of the heave exciting force impulse-response
function for a Wigley hull travelling at F, = 0.3 in following seas
where
.
2Ucos/3
-w -Ucos
. ...... .. ...............
< 2 <
---
80
4-20 Convergence properties of the heave exciting force impulse-response
function for a Wigley hull travelling at F, = 0.3 in following seas
where <go9-d
where
X
<-3oo
00
...................
.......
81
4-21 Convergence properties of the pitch exciting moment impulse-response
function for a Wigley hull travelling at F, = 0.3 in following seas where
0 <wl- <- 2Uyj81
g .
cosp
.
. . . . . .........................
.
4-22 Convergence properties of the pitch exciting moment impulse-response
function for a Wigley hull travelling at F, = 0.3 in following seas where
----2Ucos-
<W
-2 <
82
Ucos.82
4-23 Convergence properties of the pitch exciting moment impulse-response
function for a Wigley hull travelling at F, = 0.3 in following seas where
Ucosp_<
82
<
4-24 Comparison of heave exciting force impulse-response functions for a
Wigley hull travelling at different forward speeds in following seas
where 0
<
2Ucos
...
w < jj'
2UcosO. . . . . . . . . . . . . . . . . . . . . . . . . . .
84
4-25 Comparison of pitch exciting moment impulse-response functions for
a Wigley hull travelling at different forward speeds in following seas
where 0 <wl< 0
. . . . . . . . . . . . . . . . ............
. .
84
4-26 Comparison of heave exciting force impulse-response functions for a
Wigley hull travelling at different forward speeds in following seas
where cos
2U cos6' < 2 <
...........
Ucosp
12
85
2Ucos
-Ucos'
.
4-27 Comparison of pitch exciting moment impulse-response functions for
a Wigley hull travelling at different forward speeds in following seas
where
g
< W2<
g<
.......................
._ .85
_
4-28 Extended time histories of the heave exciting force impulse-response
functions for a Wigley hull travelling at F, = 0.3 in following seas. ..
86
4-29 Extended time histories of the pitch exciting moment impulse-response
functions for a Wigley hull travelling at F,= 0.3 in following seas. ..
87
4-30 Heave exciting force amplitudes for a Wigley hull travelling at F, = 0.3
in following seas.
89
.............................
4-31 Pitch exciting moment amplitudes for a Wigley hull travelling at Fn =
0.3 in following seas.
89
...........................
4-32 Heave exciting force amplitude with respect to absolute frequency for
a Wigley hull travelling at Fn = 0.3 in following seas ..........
90
4-33 Heave exciting force phase angle with respect to absolute frequency for
a Wigley hull travelling at F, = 0.3 in following seas
..........
90
4-34 Pitch exciting moment amplitude with respect to absolute frequency
for a Wigley hull travelling at F, = 0.3 in following seas. .......
91
4-35 Pitch exciting moment phase angle with respect to absolute frequency
for a Wigley hull travelling at F, = 0.3 in following seas. .......
13
91
14
Chapter
1
Introduction
The problem with which this thesis is concerned is the determination of forces and
motions of a vessel due to operation at steady forward speed in an infinite seaway.
In particular, we are interested in the ship's reactions in the presence of an ambient
wave field where body motions may be attributed to both the oncoming waves and the
scattering of these waves. This class of problem is known as the diffraction problem.
Furthermore, we are interested in developing a three-dimensional method for solving
this problem in the time domain.
Explorations for solutions to the zero speed diffraction problem have been thorough and are well documented.
The first innovations in the realm of wave force
approximations occurred in the middle and late nineteenth century with the theories
of Froude [6]in 1861 and Krilov [13]in 1896. Their method consists of integrating the
pressure due to a sinusoidal wave in the absence of a diffracting body. The pressure
is determined using a linearized Bernoulli equation, and the resulting quantity has
come to be known as the Froude-Krilov force. This force is an important component
in the overall diffraction problem as will be seen later in this text.
A further innovation in this area was presented by Haskind [8] in 1957. Here,
Haskind postulated that the overall ship motions problem could be decomposed into
separate parts. He went on to show the exciting forces (which are due to the incident
and scattered waves) may be found by the solution of a different problem, i.e., the
radiation problem.
In this case, we find the forces which are caused by radiated
15
waves occurring when the ship makes small perturbations about its mean position.
Hence the Haskind relations, as they are known, provide a means of calculating the
forces at zero speed by solving only the radiation problem. Such an approach is quite
computationally efficient and very useful when only forces for a nontranslating object
are desired.
Haskind's work was done in the frequency domain, and so in 1967 Wehausen [21]
presented analogous results in the time domain. Likewise, the fact that the Haskind
relations only apply to the case of zero forward speed begged for improvement. Thus,
in 1965 Newman [16] extended Haskind's frequency-domain relations to the case of
a body moving at steady forward speed. The forward speed relations are restricted
in their usefulness, however, since the free-surface boundary condition makes it necessary to employ reverse-flow radiation potentials which are not normally calculated.
Likewise, it must be assumed that the ship is thin. Analogous time-domain relations
are presented for the first time later in this thesis.
There has also been additional noteworthy work in the regime of time-domain
solutions to the ship motions problem. The topic was originally extensively discussed
by Cummins [3]and reinterpreted by Ogilvie [17]. In 1971, Wehausen [22]presented a
time-domain boundary integral method following that originally initiated by Volterra
[20]. This method uses an application of Green's theorem to the time derivative of the
velocity potential and an appropriate Green's function to obtain a Fredholm integral
equation of the second kind. Solution of Wehausen's integral equation provides the
velocity potential for a ship with zero forward speed. Most recently, extensions in
the theory have been made to include three-dimensional time-domain solutions for
a ship translating at steady forward speed. Such innovations may be found in King
[11], which includes ilhe diffraction problem, and Bingham et al. [2], which contains
some of the results presented in this thesis.
Most of the previous work in the area of time-domain ship motions analysis has
concentrated on solution of the radiation problem, and that which discusses the
diffraction problem spends little time with the specific case of following seas. This
thesis will attempt to clarify the subject by providing an overview of the applicable
16
theory and showing a variety of numerical results covering the main classifications of
ship speeds and wave headings, most notably results for a ship moving with steady
forward speed in following seas.
The main body of this thesis is divided into three sections. The first section,
contained in Chapter 2, illustrates the theory used in developing a linearized threedimensional time-domain boundary integral equation method for the solution of the
diffraction problem. This theory includes the simplifying assumptions that are necessary in developing the model as well as a derivation of the second kind Fredholm
integral equation and a discussion of the incident wave potential and the special considerations needed for solving the following seas case. Also, as mentioned earlier, a
derivation of the forward speed Haskind relations in the time domain is presented in
this chapter.
The second section, contained in Chapter 3, provides a discussion of the numerical aspects of solving for the diffraction potential.
Included in this discussion are
a brief overview of three-dimensional panel methods, which are used for the numerical solution of the integral equation, and a presentation of the analytical methods
with which various quantities of the incident potential and wave elevation are solved.
This chapter also contains a short explanation of how ship motions are numerically
calculated through simulation.
The third section, contained in Chapter 4, shows the numerical results obtained
by application of the theory and numerics found in Chapter 2 and Chapter 3. These
results are presented for the case of zero speed as well as for steady forward speed in
both head and following seas.
Finally, Chapter 5 provides a few concluding remarks and discussion of possible
future work.
17
18
Chapter 2
Mathematical Formulation of the
Problem
2.1
Exact Statement of Initial Value Problem
For the problem of determining the motions of a freely floating surface piercing vessel
in a seaway, we consider the model of a three-dimensional body in a semi-infinite fluid
with a free-surface. In addition, we assume the vessel translates with steady velocity
U through an incident wave field and undergoes small oscillations about its mean
position. The ship's position may be described with respect to an inertial coordinate
system, i,, which is fixed in space or with respect to a moving coordinate system, X-,
which is fixed at the mean position of the ship. The origin of the inertial coordinate
system is defined such that it coincides with the origin of the ship fixed coordinate
systemnat time t=O.
In order to evaluate the physical problem, we make the simplifying assumptions
that
he fluid is both inviscid and incompressible and that the flow is irrotational. We
can therefore describe the fluid velocity field in terms of the gradient of its potential,
that is,
V(4,t) = V$(a,,t).
19
(2.1)
Given these assumptions, the principle of mass conservation dictates that the velocity
potential is a solution to Laplace's equation
V2~ = 0
in fluid domain.
(2.2)
We may then find the pressure field, p(i,, t), in the inertial coordinate system from
Bernoulli's equation
+ gZo)
P - Pa= -p(It +
where p is the fluid density, g is the gravitational
(2.3)
acceleration, and Pa is the atmo-
spheric pressure, which is used as the reference pressure and assumed constant.
We must now define the problem boundary conditions. On the submerged portion
of the body surface, the velocity of the fluid in the direction normal to the body must
equal the velocity of the body
in V = I, -n.
(2.4)
on Sb(t)
where V(X, t) is the velocity of a point on the body, Sb(t) is the exact position of the
body's submerged surface, and n is the body's unit normal vector. The conditions
at the free-surface can be defined in terms of a kinematic and a dynamic boundary
condition. Because the free-surface is a material surface, we can state the free-surface
kinematic boundary condition in terms of the substantial derivative
D
a
Dt(Z
)+ -t ) = (t
Vs
)(uo - ) = O
on,h
o,
(2.5)
where (Z,, yo,,t) is the unknown free-surface elevation. Furthermore, we can obtain
20
the dynamic free-surface boundary condition by taking the pressure on the free-surface
as equal to atmospheric pressure:
at +
1-
2V *Vd + g
=
on z = .
(2.6)
Neglecting surface tension, we can obtain the unknown wave elevation from (2.6)
1
C(,, Yo, t) = -(Ot
g
+
1-
2VI
.
2
(2.7)
and using the substantial derivative, we may eleminate
and get the combined free-
surface boundary condition
·-,, + 2
+g
2
=
on z = .
(2.8)
Because the fluid is semi-infinite, we must also satisfy a condition as 11 -
oo.
Simply stated, for finite time periods, the velocity of the body disturbance decays to
undisturbed flow as it moves toward spatial infinity
Vt
-O
as x,y --+ oo
-40
as z -4 -00oo.
(2.9)
and
(2.10)
Finally, because the free-surface boundary condition is second order in time, it is
necessary to specify two initial conditions
21
P = O on z =O, t -- -oo
(2.11)
Pt=O on zo=O, t ---oo.
2.2 Linearization of the Forward Speed Problem
The problem as we have outlined it above is nonlinear. This is not only due to the
nonlinearities found in the free-surface boundary condition but also its application
over an unknown wave elevation (2.7). We would therefore like to simplify this problem further by linearizing the boundary conditions and Bernoulli's equation. We use
the form shown in [2] and begin by decomposing the velocity potential in the form
6
t)+Os(,t)
t)+q(X,
t)=4())+O()+E
k=1
(2.12)
where we are now using the ship-fixed coordinate system. We assume that the ship
is translating with steady speed U in the z, direction and that the flow is no longer
affected by transients due to ship acceleration. The potential as defined in (2.12) can
be expressed in terms of its steady and unsteady components. The steady components
of the potential, i(;) and O(x-),correspond to the disturbance created by the steady
forward translation of the ship. The remaining potential components are due to
unsteady motions of the fluid. The radiation potential,
k(x, t), results from unsteady
disturbances created when the translating ship is forced in a prescribed way in a
particular rigid body mode of motion. These modes coincide with the ship's six
degrees of freedom. The incident potential, q5I(x,t), corresponds to the incident wave
field through which the ship passes. Lastly, the scattered potential,
S(,t),
is due
to the scattering caused when the ship encounters the incident wave field. We will
define the diffraction potential as the sum
22
OD(z,t) =
'I(-, t) + S(-, t)
(2.13)
Returning to the decomposition of the potential (2.12), we assume that velocities
associated with the steady potential i(~) are 0(1) and all other velocities are small
perturbations about the flow described by (i).
We will call the C(1) flow the basis
flow. We require that the basis flow approaches free stream flow as we move far from
the ship but we make no other restriction on choice of t(xF). Given this fact, we
choose to use the most simple basis flow, that is, the free stream itself
(2.14)
() = -Ux.
Using this basis flow and substituting (2.12) into the expressions for the boundary
conditions and Bernoulli equation yields the Neumann-Kelvin linearization of the
initial value problem:
(2.15)
P = -P(Ot - U-)
(~at
Ua
Ox
)2 + g
9z
=0
on z
(2.16)
n:i --Un
n
(2.17)
n VIk = nkik + mkXk
23
(2.17)
where b can be
Xk,
XI,
or
s
and all body boundary conditions are applied on the
mean position of the ship surface, Sb. Likewise, we define nk as the kth component of
the generalized unit normal which for the six rigid body modes of motion are
(n1 ,n2 ,n 3) = n
( 4, n 5,n 6 ) = r x .
(2.18)
The 'm-terms' which we see in the body boundary condition for the radiation problem
are a manifestation of coupling between the steady and unsteady potentials and are
defined as
(ml, m2, m3) = -(in
(m4, m5,m6) = -(ni.
)V4(i)
V)(rx V())
(2.19)
which, given our choice of basis flow, reduce to
O, Un 3, -Un 2 ).
mk = (,0, 0,,
2.3
(2.20)
Scattered Potential Formulation
For the linearized initial value problem, we define the scattered potential, OqS(j,t), as
the solution to Laplace's equation
V 20s(, t) = 0
in V
and constrained by the following boundary conditions:
24
(2.21)
(ai- U5)qO + g =o
V gI
n -ts
.VX
=
on SF
(2.22)
on SB
(2.23)
on SO, for t finite
(2.24)
i.0
s are uniformly bounded on S, for finite t.
We should also note that both q'C and b5
Because the free surface boundary condition is second order in time, it is necessary
that we specify two initial conditions. Therefore, we also require:
O' (X-,t) -- +
t0,(X-
t) -- +
t --+- 0
(2.25)
OO
t --+
(2.26)
In deriving the integral equation for the scattered potential, we choose the time
dependent Green function derived by Wehausen and Laitone [23]
G(; t-T) = G()(;(, t-
)
G()(X-;-,t - ) = r
-
G(f)(x; .t-
)=
2 fo°
+ G(f)(; t-r)
25
(2.28)
co(tJ
kl1 - cos~-(t-
(2.27)
r)le(z+¢)Jo(kR)
(2.29)
where:
(
[(
-)2 + (y _- )2 + (Z ±)2]'
R - (X-_)2 + (y-_) 2 .
(2.30)
We begin formulation of the integral equation by applying Green's second identity
to qS(~,
r) and G(; ,t- -):
IIj(do[qs(, )V2 G(x; , t - r) = (|f
G(X; , t-r)V 2 qS(,
[is(',T )G~,(;(,t-r) -G( ;,t--
)r,(,
r)
)] = 0
(2-31)
where we describe the entire surface of integration by the sum of its parts:
S
-
SB U SF U S
U S.
We may may simplify the surface integral expression if we make note of the fact that
ff
[...]
||[I .*] =
0
2r((,t).
The former expression is true by virtue of both the boundary conditions for the
potential on So, and the nature of the Green function, which as stated earlier, satisfies
all problem boundary conditions except that for the body. We obtain latter expression
since the point of the singularity is given on the body [19]. Given substitution for the
26
above surfaces, we may write the integral in the form:
r) +s
2iroq(,
+ J
d[
T
t-)]
(;mt-a) -)G(;
(X)G
d [TS((T)GnX;(Xt-
- r)
(-,r)G(;,t-
r)] = 0.
(2.32)
In order to obtain an equation for the scattered potential itself and not its time
derivative, we integrate by parts over time from -oo to t
d
2,rXS(t) +
-
+J
I
)
drf d df
G(O)6 (t)]
-Js
GItJ
G7n -
df[s(t)Gn(o)] - ] dTJde
GT]
-
qSSGn
nGr]
=
0
(2.33)
We would like to eliminate integration over the free surface, so we begin by substituting the free-surface boundary condition into equation 2.33:
27rS(t) +
Jff()d(
[OS(t)G(O)
LidrJ[Lsc GTn
+-JLdTJs
-
- G(O)mS(t)l
SG + f/jd[S
{hi
[GTSr-
Gn[(T - 2Ud5 + d[
U2
(t)Gn(o)]
2UGrTe + U2 GTe]
E]}
0.
O=
(2.34)
Integration by parts in time of the first bracketed expression in the free-surface integral
yields
27ros(t)
+ JJ
d[OS(t)G.(O)
27
G(0)qS(t)]
-]drJ
+ -|
d [
GTr -
OnGr]
drJj d [20TGr
+ 2Gr + UGr4e- UbG] = . (2.35)
Lastly, we may apply Stoke's theorem to the free-surface integral and reduce it to
a contour integral over the waterline r to obtain the final equation
= (&)
21rns(t) +
s(t)G() + Lt drf
do
U o drTfnldl[Gf)
(2s
SG(f)
- Uo
E
0 S) + USGEd)]
t d|ff
-||
d
doG(JJS
do qGt f)
_-oOJJSb
(2.36)
where we have substituted:
Gt = -G.
2.4 Diffraction Potential Formulation
Should we wish, we may also solve for the diffraction potential directly. In order to do
so, we must first derive an integral equation for the incident potential
'(, t) which
is a solution to the interior problem, that is, valid inside the body in the volume
V. We may then use this expression along with the previous expression derived for
the scattered potential to get an integral equation for the diffraction potential. This
method has been outlined in [12].
In the body interior, e'I(:, t) must satisfy Laplace's equation
V2I(-, t) =
in V
28
(2.37)
as well as the boundary conditions
a -
aI
90
a )2
U e.V
on SF
(2.38)
on SB
(2.39)
=g. Os
n V
3~
n V ~~
and initial conditions
o'(, t) - 0
~t(,t)
-o
as t - -oo00
(2.40)
as t - -oo00
(2.41)
If we now solve for the incident potential in a manner similar to that which we used
in the previous section, we obtain
n
d G(r)
- d JJA
OI(t)Gn(
- Jjdo,
2irqA(t)
) (2 - u) + UG(
+ - dr nidl[G(C
=JjdG(O)t) - |
dr
XG(f)
(2.42)
Subtracting this from the expression for the scattered potential leads to the integral
expression for the diffraction potential
27rD(t)+ fiJ
-
U
d r j ndl[G(f) (20
= 4iS
* -
°)
D(t)G(
- uD)
] di]
29
+ UDGC()]
bGftnQ
(2.43)
2.5
The Impulse-Response Function
The most interesting quantities we may obtain by solving for the diffraction potential,
using either equation (2.36) or (2.43), are the forces on the body created by the
incident and scattered waves. As shown in [3], it is convenient to represent these
forces in terms of impulse-response functions. In so doing, we are able to predict the
exciting force response given any arbtrary incident wave elevation.
We begin by defining the pressure over the body. The pressure due to the diffraction potential can be determined from the linearized Bernoulli equation:
=
P
at + V
V )
(2.44)
We may therefore describe the related forces as the integral of the pressure over the
body
)
F=pjd( a+ V V ODnj,
(2.45)
where nj represent the components of the normal vector for the rigid-body modes
of motion as defined in equation (2.18). This expression is inconvenient, however,
since it requires that we know the gradients of the potentials as well as the potentials
themselves. We may alleviate this concern by applying Tuck's theorem, i.e.,
fJde [mm
qs + nj(V
)] = -
dOidnj (lx ni) Vt,
(2.46)
where mj represents the m-terms as defined in equation (2.19). Substitution of (2.46)
into (2.45) yields
30
Fj
pJ
-
-d
(a-
j
- Dmj)
+p dloDn3(i x Z2D) V'.
(2.47)
Given that we may define the time dependent exciting forces by a convolution of an
impulse-response function and arbitrary input wave elevation
Xj =
-oo
drKj(t - r)¢(r),
(2.48)
and we may define similar expressions for the diffraction potential and its time derivative (see [5]), the impulse response function representing the exciting forces can be
given as:
KjD(t)
-pP= I
a
j
e
d
Di
-
fj dlDcnj(lx n') VO]. (2.49)
We may assume that the waterline term is negligible if the vessel is thin and wallsided.
2.6 The Forward Speed Haskind Relations in the
Time Domain
In order to determine the exciting forces due to the scattered and incident potentials in
the ship motions problem, it has been shown that for zero forward speed, the Haskind
relations are quite useful in both the frequency domain [15] and time domain [22].
For the zero speed problem in the time domain, Wehausen was able to obtain the
expression
31
Xi(t) =
Jjd
t)qi(t,
4(d
)- 0(
+f dof ; dtt
t)i(
t -t
t
(2.50)
where bi(:i, t) is the solution for the radiation problem in mode of motion i. We can see
the advantages of using this form to obtain the exciting forces. If we were concerned
with many incident wave headings and used the scattered potential formulation, we
would have to solve as many scattered problems as there are wave headings. On the
other hand, if we use equation (2.50), we must only solve one radiation problem. It
would be nice, then, if there could be found analogous expressions for the forward
speed problem. The case of forward speed Haskind relations in the frequency domain
has been discussed in [16]. As opposed to the zero speed case, where the forward
flow radiation potential was used as assisting potential, here it is necessary to employ
the reverse flow radiation potential instead. In turning our attention to the forward
speed case in the time domain, we shall find the same to hold true.
We begin by defining the exciting force due to the scattered potential of a body
moving with steady forward speed as:
Xi(t)
= -P
c
[t
(a
t) + V'(i)
V
(d, t)]ni
(2.51)
where Ve is the velocity vector due to the steady potential. In order to obtain an
expression corresponding to Xi(t), we will make use of Green's second identity with
the scattered potential S(x, t) and the reverse flow radiation potential
(, t - r)
as the assisting potential. These potentials must satisfy the conditions set forth for
their respective boundary value problems in the moving coordinate system. For the
scattered potential, we have the familiar boundary conditions:
V 2 kS(i,t)
= 0
32
in V
(2.52)
(a - Uoa
at
ax:)2s
gSgz
a+ = 0
on SF,
(2.53)
on SB,
(2.54)
on SOOfor t finite.
(2.55)
and
The reverse flow radiation potentials must satisfy a different boundary condition on
the body and are described by:
)2-
V 2 , ( t - r) = 0
O
a
(
+U
Ot
dJ
n'*Vk
aZ
-
= nkak - mkak
in V
on SF,
on SB,
(2.56)
(2.57)
(2.58)
0k - 0
on S for t finite,
(2.59)
VO-IC --+ 0
on So, for t finite.
(2.60)
and
As seen before, we require two initial conditions for this problem which are
¢S(,
t) -O
on z = O,t - -oo
(2.61)
on z = O,t --, -oo.
(2.62)
and
q(x, t) - 0
33
Application of Green's second identity yields:
-|
d-S(1,T)O(
If
-
[4r(T)c v ((,
t T-)-
q(t-t - )V 2
k (r; t - )]
= 0
(2.63)
Therefore, we are able to construct the boundary value integral equation:
+ff )~k(
Amf[r((
tT)- (,)q(t
k -)]
+|| dft[,(, )kn(Zt - 7)- n(- -)k
+|ff
dg0[Or'(4-1O)
*m((
t-r)-
4n(gx
The integral over So gives no contribution since
nsand
(,t - )]
(t0(2.64) T'7)]
°
on approach zero as R --+ oo.
We now integrate over time from -oo to t to yield
t
._
0
Sb
+
|d dg
[ ( ,-)¢(t- )k (
(,ot-T)]
= (2.65)
If we integrate the body surface term by parts and substitute the free surface boundary
condition (noting carefully that
|Os(o
-if .
g
=-
) we arrive at the expression
k we|a
= - otrre texpresin
at
[(( r [-)r - r)- 2Uo¢e (,t
-r)+
TSf
-o
-'
(, t - )[s.(
r) - 2UoqS
34
) + U2¢,-((,r)]
0
=
( .)]
2,t
(2.66)
Let's consider only the free surface term for a moment. Separating terms by factor
of forward speed we find
)t
s1 I
2Uott t $
-- Xt dr dt [-7(
)+#4((
+ k ( t - )0((
t-a)
)]
--j drd[J(k,
7,t-r)- (,t-)S((,
)](2.67)
-)k~(
If we integrate the second half of the first term on the right hand side by parts, we
may rewrite this part of the free surface integral as
--
driL e)kA(,t
ad
(2.68)
- r)]
which is identically zero due to the initial conditions of the boundary value problem.
Recombining the right hand side gives
--i J
2U
+k(,
t
-r)]
r-)+ U20
t - )[2Uo.O.t((,r ) + vU rq(6,r )]]
(2.69)
We now integrate the second term on the right hand side by parts in order to get the
integrand of the free surface integral as a partial derivative with respect to C
Is
= --
-
-
drf d ['(, r)[-2UoP
o~ Js
o
a-
-
35
-
(, t - ) + U'(
UY(-,
4
,t-
-
7)]
t- )[UOT(g -t)+Okzer((
(2.70)
()]]
and so we may move the integral over the free surface to one over the waterlines r
and ro,
I=|L dTr d
-(- T')]]
(4 t -r)]
t - r) - Q)0-U
(2.71)
UO+ (g,
Once again, the integral over r,o gives no contribution.
Substituting the waterline
integral, the full integral equation becomes:
IJ dg S(
t)(0)
+
LdrJ
d
[OS(g r)05t(g t
r)]
f~s100kJ
o)+b-o
ft dCn( t-r)]
td
+-
['Yt
[( )[Uokt(, t +ktv(g,
t- r)[U0 tS(gr)
r) +U:(, t - )]
-
(2.72)
0XS(, T)]]
where we have substituted partial derivatives with respect to - with partial derivatives
with respect to t. Imposing the body boundary condition for the reverse flow radiation
potential we find
ff/~ *q(~ t)nk -
dTd
df
d (gt)(°) + f dfjd[
Js,
±
g
OnSbJ
jd
b
+-)
lotrt4"O['-okt(, t - 0
[qs(~,r)mk]
0(gLr)q
,t -
)]
+U ( -1
t - 7-]
++k,(t - )[vo0t (( ) - Uo (, )]]
Now we may employ Tuck's theorem [18] which, as applied to our case, states
36
(2.73)
J d4Amk
)nk
(Ve
-Jjd
=
-fdl
Snk(ilx 2D)*
(2.74)
If we make the assumption that the body is slender, the waterline integral is
of higher order than the surface integrals and so is negligible. We substitute this
expression as well as the body boundary condition for the scattered potential into the
integral equation to obtain
drIdd [Ve)n
Jdo $S(, t)nk +I|
+-
-
2
~/tdrjdi
[t(s(,-)[U0 , t - -)+
)[U - (
+kt
t(Ui
u
r)]
(r)]]
(2.75)
At this point, we should make a few comments regarding the use of the reverse
flow radiation potential. We can see that this choice of assisting potential has helped
us not only to obtain the correct term for the left hand side of our expression but
also to move the free surface integral to the waterline. The former is accomplished by
providing a suitable body boundary condition for the radiation potential. The latter
is accomplished by virtue of the free surface condition. Similar results can be seen
from the treatment of the frequency domain problem in [16]. If we once again assume
a slender body, the integral over the waterline is small and neglible. We now take a
derivitive of the expression with respect to time
d4
-
d,(4t)k()
L[Ot
\(4,
O
VaZ
Ldr
37
. VO
)]
I [tnt
bdv
-)]
(2.76)
and finally on making the substitution:
¢>t = Nit
S=
t
-D
V
we obtain the expression for the exciting force as:
Xi(t) =
=
-
d
t) +
t
-[S(-, t) + V *V0 (, r)]nk +
+X drfrOde t(
2.7
V
(, )]
Sb
)
t k
](2.77)
Special Case: Following Seas
The general expression for the linear system response due to an impulsive incident
wave field may be written as
where K(t) is the impulse response function corresponding to the incident wave field
given by the potential found in King [11]
q (~,Lt)
and
,
ig
e[k-ik[(xcos8+ysin,3]+iw,,et]
(2.79)
(t) is the input wave elevation taken in the ship-fixed coordinate system. In
general,
(t) may be given in terms of the more readily available information in
38
the inertial coordinate system if we consider the following convolution using a timevarying kernel as shown in King [11]:
r 7)(T)dT.
_(t)=
00 h(t -
(2.80)
Substitution of C(t) = e(iwt)leads to the impulse response function
h(t-Tr,)=W) { I00
| (ikUCos#pr)e(iwt)e(
-iW)dw}
(2.81)
and using the properties of convolution produces
(t) = A?{-o
a\rI
dwe 1
-
-
)
e(iet)
V2Uwcos3
}
(2.82)
where C(w) is the frequency dependent wave elevation input in the inertial coordinate
system and we have substituted
dwe= (1-
9
)dw
(2.83)
which we may obtain from the relation
We =
-
2 cosf
UW
(2.84)
We now turn our attention to the specific case of following seas. When we consider the wave elevation in following seas as measured from the ship-fixed coordinate
system, we realize that there will be an ambiguity when we try to resolve this input
in terms of its absolute frequency. The reason for this ambiguity is that there are two
39
wavelengths apparently overtaking the ship and one which the ship is overtaking, yet
all three wavelengths have the same encounter frequency. If we are measuring wave
elevation from the ship-fixed system, we have no way of di,stinguishing between the
three different wavelengths. We may reconcile this situation by considering the input
in three separate parts which fall in the restricted range of wavelengths:
O< k, < 4U2
9 2<
k2 <
4U2 cos 2 3
g
cos 2
3
g
U 2 Cos2 3
<9k < 00
2
U cos 2
3
It is interesting to note how waves in each of the three groups appear to the
ship and where the wave groups are positioned with respect to the ship. Given the
definition
we
= K-
9
where
K
II
(2.85)
is the ratio of wave number to the wave number of waves with phase speed
U cos ,3. For the case of three-dimensional
= kU
Then we have
CS2
kf4U2
=
=
atw k =
2
cos 2
9
(2.86)
.
both Ir < 1 for k <
and k =
Thus, we may relate
waves we can define re as
(2.86)
~
and Irl
at
to both the phase velocity and group velocity
of the waves:
1
P
U
40
(2.87)
Vg
1
U
2-/i
We see, then, for the range of kl, 0 <
(2.88)
< , and so the phase velocity and group
velocity of these waves are always greater than the ship velocity. For the range of k 2 ,
<
<
1, and so while the phase velocity of these waves will be greater than the
ship velocity, their group velocity will be less than the ship velocity. Finally, for the
range of k3 , c < 1, and so both the phase and group velocity of these waves will be
less than the ship velocity.
Given the ambiguity in absolute frequency, we find a need for three separate
impulse response functions and thus must modify our force convolution equation
accordingly:
3)
K
(2.89)
m=n---where now the wave elevation in the ship-fixed frame is given by
CM(t) =
R.·
1
g/4Ucosl3
{T]o
= RIIJ
e(i"t)dwe
(i)
1
-
_2uwmcosp
1gwcos
m= 1,2
m
7(r'
e(-'iwt)dw.e}m = 3
(2.90)
(2.91)
Likewise, we may obtain three separate impulse-response function representations for
VOI(, t) and Ot(zF,t). These may be given by the form found in [11]:
i cos]
7r
j sin3
Jfg/4U cos/3 we
[kz-ik[x
coso+y sin ]] e('i('t)dwe }
for m = 1,2
41
(2.92)
1
i COS
00 w
3sin P
e[kz+ik[z cos/+y
sin 3]]e(iwet)&de
}
kii
C
for m = 3
(2.93)
and
OI9(XI t);9/
o
eWle
e
m-i
6form+ysin=
1,2'dw,
for m = 1,2
= I
(,0
w e [kz+ik[
cos+
(2.94)
sinl3]]e(iet)dWe}
for m = 3
(2.95)
Finally, we can also express the frequency for each range as a function of encounter
frequency
4U cos 3,
we
9
9
2Ucos,3
9
2U cos ,
2U cos/3
1 -J
1+v/1
4U cos, 3
9
1+ 4U coS 3)
(1 +
42
(2.96)
(2.97)
(2.98)
Chapter 3
Development of the Numerical
Scheme
3.1
Panel Methods
Having laid out an analytical scheme for calculation of the velocity potential, we must
now develop a numerical scheme which will solve equation (2.36). In order to do so,
we must be able to approximate the surface of the body as well as numerically solve a
Fredholm integral equation of the second kind. We can accomplish both tasks using
a class of processes known as panel methods. In the case of the former task, we use
a number of plane quadralateral elements (triangular elements when necessary) to
describe the body surface. In the case of the latter task, we discretize the integral
equation as a system of linear equations and solve it using standard matrix techniques
such as LU-decomposition. This subsection will describe the general approach used
in this numerical technique.
The pioneering effort in this area using an arbitrary three-dimensional body was
presented by Hess and Smith [9]. The first step is to define the body by a set of points
taken along its surface. The points are then grouped appropiately to form panels.
Most often the panels are quadralaterals, however, occasionally triangular panels are
used, especially in areas of high curvature. Likewise, areas of high curvature tend to
necessitate a greater number of panels to produce accurate results. Determination of
43
the surface points depends upon the body being modelled. Generally when modelling
a ship's surface, the only available information is from its lines drawings. In this case,
the shape of the hull's cross section is known at a number of stations along the length
of the ship. Such a method of choosing surface points is conceptually simple. In a few
instances, however, it is possible to mathematically define the body (e.g. the Wigley
hull form). This procedure is also quite useful in modelling bodies such as tension leg
platforms or submarines.
At times, the four points describing a quadralateral are not coplanar. Two possible
means of rectifying the problem can be using either triangular panels or the method
described in [9]. An equivalent method is employed by TIMIT . Here, vertices are
connected using straight segments. The quadralateral plane is then defined by the
midpoints of these segments with the original vertices projected onto this plane.
Once the body has been discretized, it is necessary to discretize the integral equation. We begin by assuming that the potential is constant over each panel and use
the panel centroid as the collocation point. Thus for each panel j we can write the
approximation
O = vj.
,On
If we have the general integral equation
2 nuineaaroxn
e
G(iF;)dSt
thn G(i;u(a
u(i)dSu
then using the above approximations we can write it in its discrete form:
44
(31)
foG
N
27r~i+ E
jf
j=1
adS
N
= E vj
GdS
(3.2)
j-1
where N is the total number of panels describing the body and the surface integral is
taken over the jth panel.
Discretization of the Integral Equation
3.2
In the previous subsection, we saw the general form of the discretization of a Fredholm integral equation of the second kind. We now apply this method specifically to
equation (2.36). In this case, we are dealing with a potential which is a function of
both time and space so we must now work with discretization in two variables. Time
discretization becomes a factor in computing the convolution integrals. These convolution terms are found only on the right-hand side and may be calculated using the
trapazoidal rule. Keeping this in mind, we may determine the specific linear system
NP
2Mo
+ E
a G()(;
An
OjGMJ)
On,
j= dS
)
m=-1 j-r1
U NW
9 j=l Jj d i
3
t
M-2NW
+-9g
m=1
j=1
)
.t3
OM-
(OJ+ - OjM_~)f.,nldl(ccfM;¢
at
UNW Oj
U 31
45
Wdl(;Ott
_
nldl( Gx;0)x
IrM.-2
(3.3)
where Sj is the surface of the jth panel, rj is the waterline segment of the jth panel,
NP is the total number of panels used to describe the body, NW is the number of
waterline panels on the body, and M is the total number of time steps [5]. The prime
on the temporal summation is used to denote that a weight of one-half is applied when
m = 0. Equation (3.3) represents an NP x NP linear system with NP unknowns for
each time step.
One very nice property of this linear system of equations is that the matrix on
the left-hand side need only be factored at two time steps, the first time step and the
third time step. Investigation of equation (3.3) reveals why this is true. We first note
that the coefficients of the left-hand side matrix are essentially time indepenedent.
The key to needing two factorizations, then, lies in the use of a central difference
scheme to determine the value of the time derivative of
at the M-1 time step. We
need this value to complete the convolution in the next to last term on the right-hand
side of the equation (which in its present form only goes to the M-2 time step), but
the central difference scheme used to find it introduces two extra terms. One term,
of known quantities, is the last right-hand side expression. The other term involves
the unknown
jM and is hence on the left-hand side as the last expression. These
aforementioned terms are not present for the first two time steps, however, because
there are no convolution terms at the first time step, and the time derivative for
used in the convolution at the second time step is known to be zero by virtue of
the initial conditions. Thus, the left-hand side matrix has one set of coefficients for
the first two time steps and another set of coefficients for each time step thereafter.
The need for only two factorizations significantly reduces the computational effort
necessary to solve the problem. Also advantageous is the fact that the current value
of the unknown potential does not appear in the discretized convolution terms since
atG
O at t =
O for all 3i [5].
We can see from (3.3) that, even with the aforementioned properties, computational burden will be high when the number of panels describing the body is high. The
same is true with a large number of time steps. Thus, any means of shrinking the size
of the system while maintaining desired accuracy would be favorable. A significant
46
reduction in problem size can be accomplished by exploiting body symmetry whenever possible. Most ships are symmetric about their longitudinal centerline. Hence,
it is only necessary to discretize half the body which in turn reduces the number of
unknowns by a factor of two.
The main computational burden in the solution of the linear system of equations
comes from the calculation of the Green function coefficients. In all but the coarsest
discretizations the number of Green function coefficients is quite large (NP 2 x M) and
so it is generally impractical to store these values in memory. This being the case,
we have the option of either recalculating all the coefficients at each time step or
calculating them once and storing them on disk where they are fetched when needed.
On most workstations recalculating takes an average of five times longer so the store
and fetch method is normally preferred unless disk space is scarce. This may change,
however, as CPU speeds increase. Either way, we can approximate the necessary
storage space (RAM or disk) as
Storage ccNP 2 M.
Likewise, we can approximate computational cost as [2]
Cost oc Np 2 M 2 .
3.3
Determination of the Incident Potential
The determination of terms in (2.36) which relate to the body boundary condition as
well as those necessary to calculate the Froude-Krylov force requires that we choose
a representation of the velocity potential for the incident wave field. Because we wish
to consider an excitation which contains all possible frequencies, we must choose an
incident potential that is impulsive in nature for a chosen point (, y) on the linearized
free-surface. Given this stipulation, we may choose the incident potential of the form
47
shown in (2.79) and rewritten here
rldw<w
P'(, t)X =
g[kz-ik(cos3+ysinl)+iw.t]}
(3.4)
-
where the wave number k is a function of absolute frequency such that k(w) -= g
,3 is the angle of wave propagation, and the reference frame for the potential is the
ship-fixed coordinate system.
Using the particular choice of incident potential detailed above, we may obtain
expressions for the potential's spatial and time derivatives. In the case of head seas,
we find
'ia
(m)
VOI(X'o =
j sin s 00d(
R
2
2 U cos)
iI
e 2 [z-i(
e 9
( -X
g{
&1
cosP3+ysin3+Utcosp)]
3wU cos3 +
ir 0
9g
eit
2U2 2
92
e-[
eg [z-i(x cosP+ysin+Utcos)] eiwt
e I.
Likewise, in following seas we find
t cosP
-0I
1
2Ucose
'Qgf(,,t)
=s~
0o
2w2U cos )
9
ki
48
(3.5)
(3.6)
e
[z-i(wcos+y sin+Utcos3)]eiwt}
(3.7)
cosp
V 2(t
)=
g
. sin #
I ki
f
O(
ucomp
I-codw(
2Ucosp
2w2 U cos d
g
)
I
2 I-[z-i(mcoso+ysin3+Utcos3)le
iwt
iwt1
eo
~~~~~e
(3.8)
i cos p
V3(,t)
=
.b0 d(w - 2w2 Ucosp
dw(w )
cos
[
(9
{f 3 sin P
ki
I
e 9 [z-i(wco8s/+ysin3+Utcos3)]eiwt}
eI
ol (- t) = -9R
it xi
7r f fo
dw(l
3wU cos 3
+
2 2U2 cos2 6)
9
g2
e
02t,
w
JO
dw(l t = - -3t,
af
7r
[z-i(z cosO+ysin3+uttcos)]eiwt}
3wU cos P
9
2Ucos,6
2w 2 U2 cos2 p
92
cosP+ysin 3+Ut cos
a)]eiwt}
e e.[z-i(:
9
eI
49
(3.9)
(3.10)
J
03t(,t) =- -R
0oo
b(1-
71'
3wU cos P
9
... 'a
+
2w 2 U22 COS2
92
)
e '[z- i( cosj6+y sin 1+Utcos)]iWt
ose
1t}
(3.11)
where the subscripts 1, 2, and 3 denote the specific absolute frequency ranges over
which the integrals are taken. For the above expressions we have used (2.84) and
(2.83) to obtain integrands which are a function of absolute frequency only. It is
possible, therefore, to express all of these integrals in the form described in [11]:
I()(a, b) =
J xne-(cz 2+2bz)dx
(3.12)
where
1
a
-[-z
b
;-.2 2
+ i(x cosp + ysinp + Utcosfl)]
it
(3.13)
2
We are interested in finding numerical solutions for the expressions I(o)(a,b),
I(l)(a,b),
and I(2 )(a,b) as they apply to both head and following seas cases. A
solution for integrals of this type is given in [1] by equation (7.4.32) such that
Je-(2z+2bx+c)d
e-(f+ ~+~)d = {21- r-a
where a
$0 and
er)
+ Constant
erf(V¢Sx
+ a-) + Constant
a~~~·L
V
(3.14)
erf is the error function. If we take c = O, then by analogy we may
conclude that in general
50
e(a' 2+2b:)dx
I(°)(a, b) =
1
F
b
= ~2- ¢ ct erf(x/a + VG ) + Constant.
(3.15)
Hence, it follows that by taking partial derivatives of (3.15) with respect to a and b,
noting equation (7.1.19) in [1] which states
d= + l
d+erf(z) = (-1)2
d~~~n~~l
2
H(Z)
z
(3.16)
J~~~V
"
where
H.
Hermite polynomial of order n
Ho0=1,
we may likewise determine I(1)(a, b) and I(2 )(a, b) in terms of error functions. Beginning with I(')(a, b), we find
aJ e-(2z)dx=-2
(3.17)
and so
I(L)(a, b) = f xe(a
=
2
+2b)dx
b
Fi
2a
a
b2
-eaerff(vx
= a
+
Similarly, for I(2)(a, b) we find
51
--
e a (erf
(JI +
1
2 ae
2 e(a+2b+.)
(3.18)
J
aa
e-(a 2 +2b)dx = -
(3.19)
2e-(ax2 +2bm)dx
and so
J(2)(a
J
II)a) a f4 + ( r=e7erf(v'ax
+ib
b) =
I
=
x2e-(ax2+2bxd
-
-)
/Y±
b
/ePerf(v~
a
+
1
2
2
-
-
)+
xz
12
b2
be
-e erf(v'x+
b )e_(x2+2b+
a
b )
CO
b
(3.20)
We may now extend these general expressions to cover the specific cases of head
seas and following seas. For head seas, we wish to evaluate the integrals over the
interval [0, oo]. Noting from equation (7.1.16) in [1] that
erf(z) -, 1
as z --* oo in argzl <
,
47
and following the procedure used in [11], we obtain the expressions
I(°)(a, b) =
1)(,
1
/7I
/e
2
a [1 -
erf(
b
)1
I)( b) a2a[1- erf(b )] + 1
b) = - 2aa -e
a
VraC 2a
52
(3.21)
(3.22)
I1 b)
2 e
erf( )]
b
P)"(ci,
b = ja4CY
- ja- a-[1 -er V/a)1 2a2
%]
.
(3.23)
We now relate the error function, erf(z), with the complimentary error function,
erfc(z), using equation (7.1.2) from [1]
erfc(z) = 1 - erf(z)
(3.24)
and so we can determine the integrals of interest written recursively
(3.25)
b
I(l)(, b) = -- Io(a,b) + ca
1(2)(a,b) = -Io(a,b) -
2a
1
2a
a (b).
(3.26)
(3.27)
Finally, for implementation in the numerical scheme, we wish to calculate (3.25),
(3.26), and (3.27) in a robust manner. In order to do so, we use the relation found in
equation (7.1.3) of [1]
w(z) = e- erfc(-iz)
which gives
53
(3.28)
I(°)(ar,b) = 2 -w(
(3.29)
and is calculated using the algorithm developed by Gautschi [7].
In the case of following seas, we may determine the integral quantities in a similar
fashion; however, we now must evaluate the integrals individually over the three
specific ranges of absolute frequency. Denoting these ranges with subscripts we find:
I(°)(C' b)
J 2UcsO
e-(Q2+2b-)d
2
I r
= 1 +-e
a[erf(
2 a2U
I(O)(a, b)I
J°)uosie(a
cos
+ -)b - erf( b )]
(3.30)
aL?
+2b)dx
2Ucos
1
7 b
2
a
=[erf(/
b)-+
b U9erf(
U cos/3
G
b
+
2U cos 8
)]
b
- erf
~(gU\l(3.32)
- eW[1
1 b2
1
Similarly
l)(,
b) = 1/a
2
2
a
[f (
-
2U cos
54
i
b
V/a-
b
V/a
(3.31)
1
eb
b+)
.2 U
(4u2
2a
e
_ abi(a, b) - 1I
('
- -ee hm
2: +Ub
- )
1
(3.33)
2a
b )
= =
a +V/a
- erf(
')(aIb)I-(ab)
e--[erf(U os8
2U cos P
1 e [-(
- e
e
2a
b
-- =U2'
1'
b..
2
- /s [
--b) os
I()(a,
3 1b) = -22a /a
c
cos
2+
.(a, ' iL-
a (
j22gb
-,e e X o'ca
_Bb~i~o~
U"
-U
-o
U Cs's
+2ae~
kIb(°)(a, b) +
_-(
j
UcOs)
(3.34)
re,)]
[ - erf( g
1 a-e-W
_( Uo2 p ' V'cosp
2b + )
ie-(
-
)I+
2
a
+ U2,b)
(3.35)
2a
and
I(2)(ab)
1
b2
+2
b2
9
-e [erf(2
b2
g9
9
-
2a v a
1 b2
) - erf( )]
/o~ +
osf +
2U( ,os,8
2 - 2aUcos P
7a
)- er(f
(
2 +
b )]
V/
+b2
a2
b b
2- 2 ee
2
b2
ja22
1-IO)(a,b)
2aY
- bI()(a,b)
9
4aU cosB
55
e-(4u
*'2
Uc s )
(3.36)
I2)(c, b) =
/
:a
2 1
-a
gv'
ac[erfu
2
b
cos
2 cos
V-) - erf( 2U
3
U cosl+
9
au cos8
2
= 1 I20)(
2a
)e2
-
b
2aU cos P
a2
b) -b I)(a,
)e-
b)
a
+
+
)o2
(
2aU cos
9
2
2-e
coSiS
(3.37)
b
+ )
-e a [1- erf( o
'
U cos/6
b
b
g
2(
_up
(
4e(2C2+U
4caUcos 3
aat
e (
g2--
+ eW[1- erf(U cosV +
+ Ye a
1
UcC S +
9
W
- UcoS
a
4a
1b2
)
a2
9
b2
I2)(a, b) =
(2U cos
a29
co b (
1 2(
1
+
erf U cos,3
2 y-e~[erf(
2a2
b
a[erf
e-1
-
aU cosl
= 2 I()(a, b) -
b
a2
,2i
-U
I()(a,b)+
a
b,
g _(
2aU cos fl
)__
~b2
2gb
2 +_
co
-222,f
Ucos/
(3.38)
Again using (3.24) we can write
1
/W
b2
b
I(?)(a,b) = 2VI - a' [erfc(
2
N/-a
21 fAs
IO)(, b) = -/
[erfc(2U
g cos.
2V a
- erfc(
+
,
2U cos 3
-erf(
) - erfc(
-`
56
b
V/a
g,
+
Ucos3
(3.39)
(3.40)
I()(czb)= i ea [erfc(
± -)],
(3.41)
and using (3.28) we can rewrite all of the integrals in terms of w(z) as
ib
I(°)(a, b) =
ib
°_
g_
cosP
~~~2U
V/a-
I( 0)(a,b)=
,,/
cos/3
22U
(2
+2gbUcos/)w( gv
+ ib )
v/a
Ucos-6
I3)(cx, b) = 2
2
ib
iga
+2gbUcos)W(
[ec
(3.42)
ib
igv
/-l[e- (uc2 +gbUcos)w(
a
+)]
U os #
ar
(3.43)
(3.44)
V/a
Finally, we can find the potential expressions detailed in (3.5) through (3.11) using
the numerically obtainable integrals I(n)(a, b), i.e., for head seas
cs
=
VOI(X1
3 sin
= _'Q
3
1
Wi)(a'
b) _ 2U cos PI(2)(a,
b)) }
(3.45)
ki
[
(Ixt)=
i
R (I(O)(a, b) - 3U cos3 i(a,
ir
9
9
2U Cos2
92
and for following seas
57
b)
(2(a, b)) }
I
(3.46)
cos#
VOq(xg,t)=
R{
sinp
(I)(a,b)
2Ucs/3I))(a,
b)) }
(3.47)
ki
2a-l3U cosb)
+2U2 cos2 I(2)(, b))}(3.48)
92
for m = 1,2,3.
3.4
Calculation of the Wave Elevation
As mentioned earlier, we need to obtain the time dependent exciting force in order to
perform simulations of ship motions. This is accomplished by convolving the ship's
impulse-response function for the required mode of motion with input wave elevation
information. Because we choose our incident potential to be in the ship-fixed frame
of reference, we also must have the input wave elevation information in the shipfixed frame.
This necessity can pose a problem since it is most common to take
wave elevation measurements from a fixed point. Therefore, we must transform this
fixed point information into ship-fixed information. Theoretically, we have already
discussed the transformation process in subsection (2.7). We now turn our attention
to the problem of practical computation.
In the case of head seas, the numerical computation of wave elevation input in the
ship-fixed frame is relatively straight-forward. Since there is no frequency ambiguity
with head seas, we are left with the expression
(t=
duelR, s(v,,,),,,2ei@(3.49)
~rO
- 1 2,o~pg
58
We may then easily compute the wave elevation time history using a simple quadrature scheme such as Filon integration with no modifications to the equations.
The case of following seas is somewhat more complicated, however. We recall that
ambiguity in encounter frequency dictates the need for wave elevation information in
three distinct frequency ranges:
(t)=
I
rg/4coso
1
¢2(t =
1
4-U/o/"
71' Jo
g/4U'cos/3
{-r o
~Wj
C
-
dwI,,
be 1
-el)
-- 2Uwos
J g
C(W 2 )
1 -
(3(t) = {R lo
2U
-co
e"c'O}
(3.50)
eiwt}
(3.51)
2Uw2 cos
I
(3.52)
((W3)
_)eiwet}.
e,
(w CBC3
g
We can see that equation (3.52) may be computed in a fashion similar to (3.49), but
(3.50) and (3.51) pose a problem since their integrands become singular at the limit
w, = g/4U cos/3, a situation which becomes clear when we shstitute
equations
(2.96) and (2.97) respectively. For convenience, we reprint these relations here:
- 1-
2U cos
=
9
(1
1
-
4U cos i
9
we
4U cos) we
(3.53)
(3.54)
There are, however, two methods we can use to overcome this difficulty which are
outlined below.
In the first method, we derive expressions which contain only a square root singu59
larity and are thereby computable by standard numerical schemes. This transformation may be accomplished by rewriting the denominator of the integrand as a function
of encounter frequency using the above relations (3.53) and (3.54). If we define
9
4Ucos
WU
we can write
(l Mt=-I f{| d
u
ir
(2(t){
e
.1/Wu
dwe
}t
,ewt}.
(3.55)
(3.56)
For reasons of computational efficiency, we would like to keep the range over which
the integrand with the singular limit is evaluated as small as possible. To this effect,
we can split the range of evaluation to create a regular piece and a singular piece.
Using C(t) as an example, we find
C(t)= CR)(t) +
7X
(S)(t)
-
fO
- Awl)/1
-(&e
/Wu
J-Wvdwe
7r
1
/1
)
1e/S,
egt(3.57)
where wi is a predefined intermediate value of encounter frequency. If we now look
only at the singular part, we can deal with the square root singularity by using the
60
identity
bf
)b
f (W)d =
I.
f(b
)d
2,yf(b_ 72)
=
(3.58)
and so obtain the expression
f0
Jo~71
=
dS)(t)
oil
W= sI{
RI A-2T
-
dy7 (W - y2)ei(w-,
2
)t)}.
(3.59)
Therefore, we can compute Cl(t) using the expression
l0i
{1
h-)=
+RI
f
:' /OU
e}
J- we/wu
dy(w
Likewise, we can find a similar expression for
C2 (t) =
- 2),ei(w,,-r2)t)j.
(3.60)
2 (t):
t
eiwe}
1
dw,
af0
" r"
W./wu
d
(W2)
+
1
iF
J
W
d'7((w
-
7 2_)e'i(Wu12)t)}
(3.61)
The second and perhaps more intuitive method is to express the wave elevation
integrals entirely in terms of absolute frequency. In so doing, we are able to remove
the singularity completely and are able to obtain the wave elevation time history in
the form
1
6(0 = RI{
f
WrJ
g/2Ucos
&.olC(w, )e
61
9(-~lL))
(3.62)
1
C2(t) =
7R~I-
(
/glUcos8,
dw
I
2
d/(Uei2(lC(e
1/d
¢(t) = { 7I1,Jg/Ucos/ d3(w3)ei(
3.5
)t)}
i (1 *
0
)t)
(3.63)
(3.64)
Simulation
and hydrostatic quantities in-
One goal of determining the various hydrodynamic
volved in the ship motions problem is the ability to analyze seakeeping characteristics
through the use of simulation. In this process, we may determine the ship's movements as it steadily translates through a prescribed incident wave field. In order to
perform the simulation, we must integrate the equations of motion for the ship in
time. In obtaining the equations of motion we again assume rigid-body motion with
six degrees of freedom considering small perturbations
about the mean body position.
Given a linear system, we may then write the equations of motion in a similar fashion
to that found in Cummins [3]
6
E [(Mjk+ jk)Xk(t) + bjkh(t) + (Cj, + Cjk)k(t)
k=l
f
+
drKjk(t- r)k(r)] =Xj(t)
j
1,2... , 6,
(3.65)
where
Xj(t) = j
dTKD(t- )((T)
In this equation, Mjk constitutes the ship's mass moment of inertia matrix, Cjk is the
62
matrix of hydrostatic restoring force coefficients, jk, bk, and cjk are hydrodynamic
coefficients obtained from solution of the radiation problem, and Xj(t) represents the
time dependent exciting forces. The convolution integral on the left-hand side of
the equation, whose kernel is a product of radiation impulse-response fuctions Kjk(t)
and displacement Zk(t), is a consequence of the ship's radiated waves which, when
generated, affect the ship at each successive time step. Likewise, the right-hand side
exciting forces are represented by a similar convolution. The upper integration limit of
oo in this convolution can be attributed to the dispersive nature of the incident wave
system. We have defined the incident wave field impulsively such that it includes
waves of all frequencies which coalesce at time t = 0. Hence, some disturbance is
experienced before t = 0 due to the dispersed waves and requires that the impulseresponse function has value at times less than zero. It should be noted that this
representation of the equations of motion differs from that given by Cummins [3]
only in that the convolution for the memory function is taken with displacement as
opposed to velocity.
From numerical integration of (3.65) we may obtain time-domain histories of ship
responses in the six degrees of freedom. These time histories may be validated if we
compute frequency-domain response amplitude operators (RAOs) from the input and
output signals:
RAOj=
IX(,)1
(3.66)
X j (w) and (w) are determined by taking the Fourier transforms of the output and
input signals respectively.
63
64
Chapter 4
Numerical Results
The numerical scheme discussed in chapter 3 has been incorporated into the suite of
programs called TIMIT . As well as solving for the exciting force impulse-response
functions for the six rigid-body modes of motion, these programs evaluate the time
histories of the canonical potential values on each panel for both the radiation and
diffraction problems, determine the hydrostatic and hydrodynamic coefficients, and
solve for the radiation problem impulse-response funtions [5]. All input and output
quantities are suitably non-dimensionalized as explained below:
Ship Speed: U* =-
Spatial Dimensions:
Time: t
t
-
65
=
Incident
Wave
M ,lU
X
Figure 4-1: Definition for the incident wave angle heading.
Impulse-responseFunction: Kf
-
D
K/i
3D
pCdL
kg/
k = 3 for j = 1,2,3
k = 4 for j = 4,5,6.
Hereafter, the stars will be dropped and the non-dimensionalized form implied unless
otherwise stated. A graphical representation for the definition of the wave heading
angle may be found in Figure (4-1). Thi- definition means that for head seas
and for following seas /3= 0. Straight beam seas are given when / =
= r
.
All numerical results presented here were obtained using an IRIS Indigo workstation. Both spatial and temporal discretizations were as coarse as possible while
maintaining converged results. The numerical experiments were run using a Wigley
66
hullform with non-dimensional characteristics as follows:
Length =1.0
Beam =0.1
Draft = 0.0625
Midship Section Coefficient = 0.909.
This hullform corresponds to Wigley model 1 which was used to obtain experimental
data by Journee [10].
4.1
Zero Speed Results
Zero speed calculations are merely a special case of the forward speed model presented
in section 2.3. A quick review of equation (2.36) shows that the waterline integral
vanishes for U = 0 and so there are no waterine effects. The zero speed results are
interesting in as much as they are readily verifiable. There are some experimental
data available as well as a number of reliable computer programs which solve the zero
speed seakeeping problem in the frequency domain. Hence, comparison with existing
results provides a good initial verification for the numerical model.
Figure (4-2) shows three snapshots in time of the typical incident wave profile
encountered by the stationary vessel - one before the waves have coalesced, one at the
point of coalescence, and one after the waves have begun to disperse. As is pointed
out in [2], these profiles have actually had a small amount of the short wavelength
contribution filtered out since we are evaluating
67
0
0.75
0.50
t =
O.
0.25
-0.25
-0.50
-0.75
-2 0
-10
0.0 X
1.
2.0
Figure 4-2: Profiles for an example impulsive incident wave given F, = 0.0 and f = r
[2].
(x, t) -
(z, y, -0.001, t)
and the waves are attenuated exponentially as we move away from the free surface.
Because we evaluate the potentials at the panel centroids, the filtering of short waves
is present in the results. This effect is not entirely detrimental, however. The fact that
we approximate the body by plane panels and the potentials by piecewise constants
means that we are unable to resolve wavelengths which are not substantially longer
than the panel dimensions. The presence of very short wavelengths can then cause
numerical errors, although they are irrelevant in most cases. Hence, the filtering can
help remove these anomolies.
Impulse-response functions for the heave exciting forces and pitch exciting moments of a stationary Wigley hull in head seas are shown in Figures (4-3) and (4-4)
respectively. Because the Wigley hull form is a nice shape that provides no computational difficulties (i.e., it is wall-sided, has no flair, etc.), relatively few panels are
68
0.040
K 3D
0.030
0.020
0.010
0.000
-10
-5
0
5
t
10
Figure 4-3: Heave exciting force impulse-response function for a Wigley hull travelling
at F, = 0.0 in head seas.
needed to obtain converged results. In this case, converged results were obtained with
a panelization of 144 panels over the half-body exploiting port/starboard
symmetry.
For verification purposes, these results were converted to the frequency domain
by the Fourier transform
Xj(w)r=
t
dtKjD(t)ei-'
(4.1)
where the transform has been numerically calculated using Filon quadrature based
on equations (25.4.47) through (25.4.56) in [1]. For this quadrature scheme, a time
step corresponding to converged impulse-response function results is used (At
0.1) and equations (25.4.47) and (25.4.54) have been truncated to exclude terms of
order At5 and higher. Figure (4-5) shows the magnitude of the frequency dependent
heave exciting force and likewise Figure (4-6) shows the magnitude of the frequency
dependent pitch exciting moment. Both sets of results are compared with similar
quantities calculated by WAMIT, a frequency domain analog of TIMIT [4]. We can
69
K5D
0.0030
0.0020
0.0010
0.0000
-0.0010
-0.0020
-0.0030
-10
-5
0
5
t
10
Figure 4-4: Pitch exciting moment impulse-response function for a Wigley hull travelling at Fn = 0.0 in head seas.
see the results are identical to graphical accuracy and so we may be fully confident
with our numerical scheme for zero forward speed.
4.2
Forward Speed Results: Head Seas
Turning our attention to the more interesting case of a body in steady forward translation, we first look at this body in head seas. Time-domain results for the diffraction
problem in head seas at forward speed have previously been shown by King [11].
There is also an existing pool of experimental and numerical data in the frequency
domain. We will use this data to further verify our model.
Figures (4-7) and (4-8) illustrate convergence properties for typical heave and
pitch diffraction impulse-response functions respectively. We see the results for three
separate discretizations of a Wigley hull translating at a Froude number of 0.3. As is
exemplified by these figures, it is possible to obtain practical results with a panelization of 144 panels over the half-body exploiting port/starboard
convergence occurs for a time step of 0.1.
70
symmetry. Temporal
X/pg;,(AL)
20
15
10
5
0
0.50
1.00
1.50
Figure 4-5: Heave exciting force amplitude for a Wigley hull travelling at F = 0.0
in head seas [2].
X/pDRC
,
.
.
0.50
1.00
(iA
1.50
Figure 4-6: Pitch exciting moment amplitude for a Wigley hull travelling at F, = 0.0
in head seas [2].
71
K3D
0.050
0.025
0.000
-10
-5
0
5
10
Figure 4-7: Convergence properties of the heave exciting force impulse-response function for a Wigley hull travelling at F, = 0.3 in head seas [2].
KSD
0.0050
0.0000
-0.0050
-0.0100
-10
-5
0
5
10
t
Figure 4-8: Convergence properties of the pitch exciting moment impulse-response
function for a Wigley hull travelling at F,, = 0.3 in head seas [2].
72
A noteworthy aspect of the impulse-response functions shown in Figures (4-7) and
(4-8) is the presence of non-zero oscillations for time t < 0. This characteristic, also
seen in the zero speed results, is due to the nature of the prescribed incident wave
field. We saw earlier in section 3.3 that the incident wave field we have chosen is an
amalgamation of all possible wave frequencies and is defined such that all of these
waves come together at the ship's origin at time t = 0. As a consequence, the bow
of the ship passes into the wave field for t < 0 while there is still some dispersion.
Hence, the bow experiences some excitation before t = 0 and these excitations show
up as small pre-peak oscillations.
The effect discussed above may best be illustrated by Figures (4-9) and (4-10).
Here we see both heave exciting force and pitch exciting moment impulse-response
functions for different levels of forward speed, the fastest being at a Froude number
of 0.4 and the slowest being zero speed. Note that as the ship moves at faster forward
speeds, the peak becomes higher and sharper with the onset of pre-peak oscillations
coming at a time closer to t = 0. This phenomenon is a manifestation of the ship
moving in the opposite direction to the incident wave field and the fact that we
have defined the incident potential with respect to the ship-fixed coordinate system.
Because the oncoming wave field is timed to coalesce at
= 0 and the wave field is
described by the encounter frequency, the bow of the faster moving ship arrives when
the waves are less dispersed and hence feels less of the pre-peak effect. Likewise, the
more complete or "tighter" coalescence results in the ship experiencing more energy
closer to the peak.
We may obtain frequency-domain results from our impulse-response functions by
making use of the Fourier transform:
Xj(W)=
o
dtKjD(t)eiWet
(4.2)
where the exciting forces (or moments) are now a function of encounter frequency.
These forces are most useful to us as a function of absolute frequency and may
73
K3D
0.05
0.02
0.00(
t
Figure 4-9: Comparison of heave exciting force impulse-response functions for a
Wigley huli travelling at different forward speeds in head seas.
K5D
0.0050
.
----
Ii---I -. /'
0.0000
I,
--
Fn=O.O
--
0.2
Fn=0.4
%i1
I
%I I I
'i
-0.0050
-0.0100
.
-10
~
~
-5
~
.
.
.-
.
0
.
.
.
5
t
10
Figure 4-10: Comparison of pitch exciting moment impulse-response functions for a
Wigley hull travelling at different forward speeds in head seas.
74
X/pg,(a/L)
(LA)"2
Figure 4-11: Heave exciting force amplitude for a Wigley hull travelling at F, = 0.3
in head seas [2].
easily be converted as such given the relation in (2.84). Figures (4-11) and (4-12)
show the magnitude of frequency-domain exciting force and pitch exciting moment
for a Wigley hull at a Froude number of 0.3. Similar plots for phase angle may
be found in Figures (4-13) and (4-14). In both sets of plots, TMIT
results are
compared with experimental data of Journee [10] and numerical results from the
Rankine panel method SWAN [14]. Agreement in the case of heave exciting force,
both magnitude and phase angle, is quite good. The results for pitch exciting moment
amplitude are not as nice, however. The reason for the discrepencies may lie in the
method of the calculation of the m-terms which affect the force computation as seen in
equation (2.49). While TIMIT uses a Neumann-Kelvin linearization, SWAN employs
a linearization about the double-body flow. Because the value of the m-term vanishes
for heave but is nonzero for pitch, the difference in linearization schemes may explain
the better agreement with experiment for pitch demonstrated by SWAN [2].
The final set of results presented for head seas are those obtained from the simulation of a Wigley hull moving through a seaway at a Froude number of 0.3. The wave
spectrum used is a Pierson-Moskowitz corresponding to a wind speed of 5 meters
per second. A portion of the time history for the input wave elevation along with
75
X.11)RCA
2.0
1.5
1.0
0.5
oin
0.5
1.0
(L&)"=
2.0
1.5
Figure 4-12: Pitch exciting moment amplitude for a Wigley hull travelling at F, = 0.3
in head seas [2].
Phase
150
100
50
0
-50
-100
-150
0.50
0.75
1.00
(L,2
1.25
1.50
Figure 4-13: Heave exciting force phase angle for a Wigley hull travelling at Fn = 0.3
in head seas [2].
76
Phase
150
100
50
0
-50
-100
.1 50
0.50
0.75
1.00 (L)A
1.25
1.50
Figure 4-14: Pitch exciting moment phase angle for a Wigley hull travelling at Fn =
0.3 in head seas [2].
heave and pitch responses are shown in Figure (4-15). As previously mentioned, we
can obtain frequency-domain response amplitude operators by using equation (3.66).
Response amplitude operators corresponding to the input and output in figure (4-15)
may be found in Figures (4-16) and (4-17). Plotted along with these results are ex-
perimental data from Journee [10]. One notable characteristic of these plots is the
slow convergence around the resonant peak. This effect may be attributed to two numerical phenomena involving the solution of the discrete equations of motion - poor
conditioning of the linear system in the region of resonant response and sensitivity in
time step size for the numerical integration scheme [2].
4.3
Forward Speed Results: Following Seas
Finally, we look at the more complex situation of a body in steady forward translation
encountering following seas. Experimental time-domain results for this case are not
readily available. Likewise, frequency-domain experimental data are in short supply
so verification of our numerical results is difficult and we must rely a great deal on
intuition.
77
0
000
420
I ·
_·I
I
I ·
. . .
)
. . .
X3
3I
305
310
315
320
32J
330
335
340
345
350
355
350
365
370
375
310
30S
390
395
i 0
I
I1I
X
5°o o'
-OC
-tO
Figure 4-15: Sample data for the simulation of a Wigley hull travelling at F = 0.3
in head seas. The top plot is the input incident wave field in the ship-fixed frame,
the middle plot is the heave response, and the bottom plot is the pitch response [2].
RA(
(IA)' 2
Figure 4-16: Response amplitude operator for the heave exciting forces of a Wigley
hull travelling through a Pierson-Moskowitz wave spectrum at F, = 0.3 and = r
[2].
78
RAC
0.25
0.50
0.75
1.00
1.25
1.50
(LA/2
Figure 4-17: Response amplitude operator for the pitch exciting moments of a Wigley
hull travelling through a Pierson-Moskowitz wave spectrum at F, = 0.3 and 3 = r
[2].
Let us first examine the typical convergence behavior for the time-domain impulseresponse functions. We recall that for following seas we need three separate impulseresponse functions for each mode of motion to account for the ambiguity in absolute
frequency. Figures (4-18), (4-19), and (4-20) show heave impulse-response functions
for three discretizations of a Wigley hull translating at a Froude number of 0.3. Similar plots for pitch may be found in Figures (4-21), (4-22), and (4-23). The most
interesting trend we see in these figures is the generally fast convergence for the
impulse-response functions that relate to the two lower frequency ranges of incident
wave input (Figures (4-18), (4-19), (4-21), and (4-22)) as contrasted to the poorer
convergence found in the higher frequency range results (Figures (4-20) and (4-23)).
The relatively slow convergence found in this range of high frequencies may be attributed to the high frequency wave effects discussed earlier. More specifically, as the
body discretization becomes finer (and consequently the panels become smaller), the
effects of the smaller wavelength high frequency waves are better represented in the
numerical results.
In the previous section, we saw that for head seas the peak of the impulse-response
79
K1 3 D
0.0150
0.0100
0.0050
0.0000
-10
-5
0
5
t
t
10
Figure 4-18: Convergence properties of the heave exciting force impulse-response
function for a Wigley hull travelling at F = 0.3 in following seas where 0 < w <
2U
2Uos O'
K2 3 D
0.00150
0.00100
0.00050
0.00000
-0.00050
-0.00100
-10
-5
0
5
t
10
Figure 4-19: Convergence properties of the heave exciting force impulse-response
function for a Wigley hull travelling at F, = 0.3 in following seas where
a , <
Og2
lUcos3
U2
<
r cos,6'
80
K3 ,,
0.0075
0.0050
0.0025
0.0000
-0.0025
-0.0050
-0.0075
-10
-5
0
5
t
10
Figure 4-20: Convergence properties of the heave exciting force impulse-response
function for a Wigley hull travelling at F = 0.3 in following seas where -Uc <
W3 < 00.
Kl 5 D
1.5000E-4
1.0000E-4
5.0000E-5
5.4570E-12
-5.0000E-5
-1.0000E-4
-1.5000E-4
-10
-5
0
5
t
10
Figure 4-21: Convergence properties of the pitch exciting moment impulse-response
function for a Wigley hull travelling at F, = 0.3 in following seas where 0 < w1 <
2U cos
'
81
K 2 5D
0.00050
0.00000
-0.00050
0
-10
5
-
tt
10
Figure 4-22: Convergence properties of the pitch exciting moment impulse-response
<
function for a Wigley hull travelling at F = 0.3 in following seas where W2 < Ucosi'
a
K 3 5D
0.00150
0.00100
0.00050
0.00000
-0.00050
-0.00100
-0.00150
-10
-5
0
5
t
10
Figure 4-23: Convergence properties of the pitch exciting moment impulse-response
function for a Wigley hull travelling at F = 0.3 in following seas where g < W3 <
00.
82
function became higher and narrower as the ship's speed increased. One might then
expect to find an analogous trend as ship speed increases in following seas. This is
indeed the case. We can see from Figures (4-24) and (4-25), which are comparisons of
heave and pitch impulse-response functions for the input frequency range of 0 < w1 <
U
that as Froude number increases, the peak spreads out more and becomes
lower. The explanation for this behavior is similar to that found for head seas, but
here the situation is reversed. The ship is now travelling within an incident wave field
containing one frequency range of waves with group speed greater than ship speed
and two frequency ranges of waves with group speed less than ship speed(see section
2.7 for more detail). Given this condition, the faster the ship travels, the longer it
will take for the overtaking wave groups to reach the vessel's origin and point of wave
coalescence. Likewise, as the ship's speed increases, there will be fewer wave groups
which fall into the overtaking category. Hence, the stern of the ship experiences
more pre-peak response and for similar reasons the bow experiences more post-peak
response. The end result is that the energy is more broadly spread around the zero
time region.
Conversely, we see a seemingly opposite effect in the peak area for the two frequency ranges of wave groups which travel slower than the ship's speed, i.e., the wave
frequencies described by w 2 and
3
. Illnthis case, the ship is catching up with the
wave groups in these frequency ranges and so the faster the ship travels, the quicker it
should overtake them. There will be more wave groups to overtake, however. Thus we
see greater peak responses in these ranges for faster ship speeds, but we also see some
of the broadening found in the first range response. Figures (4-26) and (4-27), which
are comparisons of heave and pitch impulse-response functions for the input frequency
range of
g
< w2
<
,
illustrates these points, although the response for this
frequency range is dominated by r =
4
effects.
As mentioned above, one of the most striking features of the figures previously
presented is the magnitude of the oscillations occurring both behind and ahead of
the peak. We may attribute these oscillations to the r =
be noted that for head seas, the r =
4
wave frequency. It should
effects for the Wigley hull are minimal since
83
K1.
0.0300
0.0250
0.0200
0.0150
0.0100
0.0050
0.0000
-0.0050
-10
-5
0
5
10
Figure 4-24: Comparison of heave exciting force impulse-response functions for a
Wigley hull travelling at different forward speeds in following seas where 0 < w1 <
9
2U cos, '
K.lDU
if ·
0.0020
F
._
.i
.
.
X~~
.
.
,
.
_
-
w- -
_
-
F
--I
0.0010
I
'~.% .,-'
t
\~~~~~~ \
\
-.I I
Fn=0.3
'
I
\
I
I.
\
t
t
-0.0010
.
.
.
.
..
.
/\I)I
/
!
'\
.
...
-5
r. *
I
.
t
I.
0
s
~~~~I
I
, I
II'\
II
t
I
.
.
'
'''.
t
I
I
r~~~~~~~~~~~~~
-1 I0D
Fn0.1
8
s
0.0000
-0.0020
ww
DII
J~~~~~
._
5
t
10
Figure 4-25: Comparison of pitch exciting moment impulse-response functions for a
Wigley hull travelling at different forward speeds in following seas where 0 w
9
2Ucos
3'
84
K2 3D
0.00200
0.00150
0.00100
0.00050
0.00000
-0.00050
-0.00100
-5
-10
0
5
t
10
Figure 4-26: Comparison of heave exciting force impulse-response functions for a
Wigley hull travelling at different forward speeds in following seas where
ge <
w2
cos-
-wt~~~~~~~~~~~~~~~~~~~2U
g
- UcosB'
K2 5D
0.0005C
0.00000
-0.00050
-10
-5
0
5
t
10
Figure 4-27: Comparison of pitch exciting moment impulse-response functions for a
Wigley hull travelling at different forward speeds in following seas where
9
<
w2
y
--rrcosO'
-
~~~~~~~~~~2Uos
85
KV
'm3D
.I
F
-100
.
.
I I
-50
I
I -
I
0
. I I I
50
.
I .I
I
100
Figure 4-28: Extended time histories of the heave exciting force impulse-response
functions for a Wigley hull travelling at F, = 0.3 in following seas.
this hull form is a very weak scatterer, i.e., the scattering effects of the Wigley hull in
head seas is an order of magnitude smaller than the Froude-Krylov forces created by
the incident wave field. In following seas, however, the nature of the incident wave
field and the definition of where and when it coalesces result in the ship always riding
on the wave which has the same group speed as the ship's forward velocity. We recall
from section 2.7 that this group speed corresponds to the frequency where the r =
singularity occurs. Therefore, the ship always experiences a response associated with
this wave group - a response which grows as the wave groups get closer together and
shrinks as the wave groups disperse.
The short time duration shown in the earlier figures does not allow us to understand the complete behavior of the r =
4
oscillations. Thus it is necessary to
examine a substantially longer time history of the impulse-response function. Such
time histories are presented in Figures (4-28) and (4-29) for heave and pitch respectively. Bingham et al. [2] shows that for forward speed, the leading order of the
Green function is described asymptotically by the expression
86
KmsD
0.00075, .
0.00050
I
RmgeI
I'i
Range 2
Range 3
.....-
0.00025
0.00000
-0.00025
-0.00050
!
-0.00075
L
. ....J
I
-100
....
-50
I
....
0
!
50
....
I ·
t100
Figure 4-29: Extended time histories of the pitch exciting moment impulse-response
functions for a Wigley hull travelling at F, = 0.3 in following seas.
Gt
where w, =
k
U--e'
sin (wt),
(4.3)
is the critical frequency of oscillation. This behavior leads to the
postulation that the
T-1
oscillations should also decay proportional to l. A cursory
examination of the data presented in Figures (4-28) and (4-29) shows this to be true.
One aspect of these oscillations which may not be intuitively obvious is that the
contributions from the wl frequency range and the w2 frequency range appear to be
out of phase by 180 degrees. A closer look at the equations governing the forcing of
this problem, i.e, the spatial and temporal derivatives of the incident potential, shows
us why the oscillations behave as such. In particular, let us examine the expressions
for
VyOq
and XV7 found in equations (3.7) and (3.8) respectively.
We note that
in the case of Vol, the r =1 14 frequency contributes to the quantity as the upper
limit of integration. In the case of VI, however, the r =
-fr-quency contributes
to the quantity as the lower limit of integration. We should ilso remember that
87
at this frequency the integrands in the equations for VfI
and _Vi are the same.
Thinking only in the mathematical terms of evaluating these integrals, it should then
become apparent that the contribution of the r =
frequency in equations (3.7)
and (3.8), which begins to dominate at times far from t = 0, should be of opposite
sign and similar magnitude.
This reasoning is born out further by the analytical
expressions for these quatities found in section 3.3, however, it should not imply that
these expressions cancel each other.
As with the case of head seas, we may obtain results for following seas in the
frequency domain by taking Fourier transforms of the individual impulse-response
functions. These transforms are represented as
Xmj(Wc)=
L
dtKjD(t)esiW
t
(4.4)
for m = 1,2,3,
where m corresponds to the three distinct input frequency ranges. Again, the exciting
forces or moments are a function of encounter frequency. These frequency-domain
results are shown for heave and pitch in Figures (4-30) and (4-31). Because input
for the impulse-response functions corresponding to the first and second frequency
ranges (m = 1 and 2) only has frequency content 0
w
< <
9
we expect to
see no response beyond this region for these functions and this is indeed the case.
Likewise, because input for the impulse-response function corresponding to the third
frequency range (m = 3) contains all encounter frequencies, we are not surprised to
see response across the entire spectrum.
It is possible to obtain frequency-domain results with respect to absolute frequency
by using the relations found in eqations (2.96), (2.97), and (2.98). Such results are
shown in Figures (4-32) and (4-33) for heave and Figures (4-34) and (4-35) for pitch.
88
I
X/pg,.(4L
0.050
0.025
0..0
0.5
1.0
1.5
ci)
2.0
2.5
Figure 4-30: Heave exciting force amplitudes for a Wigley hull travelling at F, = 0.3
in following seas.
X/pg"A
0.004
0.003(
0.002C
0.0010
U.0
0.5
1.0
1.5
2.0
oi,
Figure 4-31: Pitch exciting moment amplitudes for a Wigley hull travelling at F =
0.3 in following seas.
89
X/pgUA(/L)
0.050
0.025
0.0
0).5
1.0
(If.'2
1.5
2.0
Figure 4-32: Heave exciting force amplitude with respect to absolute frequency for a
Wigley hull travelling at F, = 0.3 in following seas.
Phase
150
100
50
0
-50
-100
-150
0.0
0.5
1.0
(ILA)
Figure 4-33: Heave exciting force phase angle with respect to absolute frequency for
a Wigley hull travelling at F, = 0.3 in following seas.
90
X/p.&
0.0040
0.0030
0.0020
0.0010
0.0
0.5
1.0
(IA) "2
1.5
2.0
Figure 4-34: Pitch exciting moment amplitude with respect to absolute frequency for
a Wigley hull travelling at F, = 0.3 in following seas.
Pha
0.0
0.5
(LA),,
1.0
Figure 4-35: Pitch exciting moment phase angle with respect to absolute frequency
for a Wigley hull travelling at F, = 0.3 in following seas.
91
92
Chapter 5
Conclusions
In this thesis, we have achieved the objective of solving for the forces created by
the diffraction problem in general, and in particular, have extended the problem for
the special case of following seas. The approach used here is based on linear timedomain theory and solves for the velocity potential due to an incident wave field and
its subsequent scattering by the body. The forces on the ship are then expressed in
terms of impulse-response functions which allow us to determine the ship's reaction
due to an arbitrary wave elevation. The theory is applicable to both the zero speed
and forward speed cases.
Because we are using linearized potential flow theory, we must bear in mind that
several simplifying assumptions have been made. By virtue of the potential flow,
we have assumed that the fluid is incompressible and inviscid, and that the flow is
irrotational.
Likewise, by assuming that the problem is linear, we have neglected
the higher order terms in all boundary conditions.
As long as we consider only
small perturbations about the body's mean position, these assumptions are valid and
produce accurate results.
Much of the theory presented in this thesis is known from previous work in the
field of ship motions. An interesting new piece of theory, however, is found in the presentation of the forward speed Haskind relations in the time domain. These relations
are analogous to their frequency-domain counterparts and carry the same drawback
of having to assume that the body is thin.
93
While this characteristic reduces the
practical uses of these relations, they are still of scientific interest.
The need for modification of the general theory when solving the forward speed
diffraction problem in following seas has also been discussed. We determined that
given an impulsive incident wave field described by a potential with respect to the
ship-fixed coordinate system, we must solve for three impulse-response functions
which correspond to three distinct ranges of absolute wave frequency. This modification must be made because of the ambiguity in encounter frequency that arises
when the ship is travelling in following waves.
The complexity of the overall problem dictates that we cannot solve for the potential analytically. It is therefore necessary to develop a robust numerical scheme for
the purpose of calculating the potentials and impulse-response functions. To do so,
we have relied on a three-dimensional panel method with constant source strengths.
The second kind Fredholm integral equations have been discretized in space and time
to produce a set of linear equations for the velocity potentials of the body panels
at each time step. These potentials may then be used in the linearized Bernoulli
equation and integrated over the body to determine the impulse-response functions.
Numerical results produced by this scheme show excellent agreement with experimental results when such results exist. In the case of following seas, for which
experimental data is not readily available, the results follow intuition. In all cases it
is possible to get analogous frequency-domain information from the impulse-response
functions by taking their Fourier transforms. We may also use these impulse-response
functions, along with impulse-response functions obtained from solution of the radiation problem, to predict the ship's motions in an arbitrary wave field via simulation.
An obvious next step in this project is to experimentally obtain data for a ship
travelling at forward speed in following waves. Unfortunately, reliable experimental
results for such a case are not easily found. Other areas that merit further consideration include choice of incident wave potential and nonlinearization of the problem.
The former area has been discussed to some extent in [2]. The latter area may be
addressed partially by using a 'body-nonlinear' approach, i.e., solving the problem
using the exact body boundary conditions at each time step. Both areas may prove
94
to be excellent topics for future research.
95
96
Bibliography
[1] M. Abramowitz and I. A. Stegun. Handbook of Mathematical Functions. Dover
Press, New York, 1964.
[2] H. B. Bingham, F. T. Korsmeyer, J. N. Newman, and G. E. Osborne.
The
simulation of ship motions. In 6 th Intl. Conf. on Numerical Ship Hydro., 1993.
[3] W. E. Cummins. The impulse response function and ship motions. Schiffstecknik,
9:101-109, 1962.
[4] Department of Ocean Engineering, Massachusetts Institute of Technology, Cam-
bridge. WAMIT; A radiation-diffractionpanel program for wave-body interactions, 1988.
[5] Department of Ocean Engineering, Massachusetts Institute of Technology, Cambridge, MA. TIMIT,
[6] W. Froude.
1993.
On the rolling of ships.
Transactions of the Institute of Naval
Architecture, 2:180-229, 1861.
[7] W. Gautschi. The complex error function.
Collected Algorithms from CACM,
1969.
[8] M. D. Haskind.
The exciting forces and wetting of ships in waves. Izvestia
Akademii Nauk SSSR, Otdelenie Tekhnicheskikh Nauk, (7):65-79, 1957.
[9] J. L. Hess and A. M. O. Smith. Calculation of nonlifting potential flow about
arbitrary three-dimensional bodies. Journal of Ship Research, 8(2):22-44, 1964.
97
[10] J. M. J. Journee. Experiments and calculations on four Wigley hullforms. Technical Report 909, Delft University of Technology, Ship Hydromechanics Laboratory, Delft, The Netherlands, 1992.
[11] B. W. King. Time-domain analysis of wave exciting forces on ships and bodies. Technical Report 306, The Department of Naval Architecture and Marine
Engineering, The University of Michigan, Ann Arbor, Michigan, 1987.
[12] F. T. Korsmeyer, J. N. Newman, and G. E. Osborne. The linear, transient,
free-surface wave diffraction problem. In preparation, 1994.
[13] A. Kriloff. A new theory of the pitching motion of waves, and of the stresses
produced by this motion. Transactions of the Institute of Naval Architecture,
37:326-368, 1896.
[14] D. E. Nakos and P. D. Sclavounos. Ship motions by a three dimensional Rankine
panel method. In Eighteenth Symp. on Nav. Hydro., Ann Arbor, Michigan, 1990.
[15] J. N. Newman. The exciting forces on fixed bodies in waves. Journal of Ship
Research, 6(4):10-17, 1962.
[16] J. N. Newman. The exciting forces on a moving body in waves. Journal of Ship
Research, 9(3):190-199, 1965.
[17] T. F. Ogilvie. Recent progress toward the understanding and prediction of ship
motions. In The Fifth Symposium on Naval Hydrodynamics, pages 3-128, Bergen,
1964.
[18] T. F. Ogilvie and E. O. Tuck. A rational strip theory for ship motions, part
1. Technical Report 013, The Department of Naval Architecture and Marine
Engineering, The University of Michigan, Ann Arbor, Michigan, 1969.
[19] J. J. Stoker. Water Waves. Interscience Publishers, Inc., New York, 1957.
[20] V Volterra. Sur la theorie des ondes liquides et la methode de Green. J. Math.
Pures Appl., 13:1-18, 1934.
98
[21] J. V. Wehausen. Initial value problem for the motion in an undulating sea of
a body with fixed equilibrium position.
Journal of Engineering Mathematics,
1:1-19, 1967.
[22] J. V. Wehausen. The motion of floating bodies. Ann. Rev. Fluid Mech., 3:237268, 1971.
[23] J. V. Wehausen and E. V. Laitone. Surface waves. In Handbuch der Physik,
pages 446-778. Springer, 1960.
99
Download